Article
Reference
Playing to the gallery: Investigating the normative explanation of ingroup favoritism by testing the impact of imagined audience
IACOVIELLO, Vincenzo, SPEARS, Russell
Abstract
The present research examined the role of social norms as a determining source of ingroup favoritism in minimal groups. Across three studies (total N = 814), results showed that ingroup favoritism was reduced when participants imagined the reaction of an external (and egalitarian) entity, as compared to a control condition or a condition in which they were explicitly asked to imagine the reaction of ingroup members. In line with the prediction that the desire to appear as a good group member drives conformity to the ingroup norm, the findings also revealed that favoring the ingroup resulted in higher self-esteem (Study 2). This was however limited to situations where the ingroup norm was inferred or induced to be pro-discriminatory, but not when it was antidiscriminatory (Study 3). The proposed explanation is discussed in the light of dominant explanations of ingroup favoritism.
IACOVIELLO, Vincenzo, SPEARS, Russell. Playing to the gallery: Investigating the normative explanation of ingroup favoritism by testing the impact of imagined audience. Self and Identity , 2021
DOI : 10.1080/15298868.2021.1933582
Available at:
http://archive-ouverte.unige.ch/unige:155041
Disclaimer: layout of this document may differ from the published version.
1 / 1
Full Terms & Conditions of access and use can be found at
https://www.tandfonline.com/action/journalInformation?journalCode=psai20
Self and Identity
ISSN: (Print) (Online) Journal homepage: https://www.tandfonline.com/loi/psai20
Playing to the gallery: investigating the normative explanation of ingroup favoritism by testing the impact of imagined audience
Vincenzo Iacoviello & Russell Spears
To cite this article: Vincenzo Iacoviello & Russell Spears (2021): Playing to the gallery:
investigating the normative explanation of ingroup favoritism by testing the impact of imagined audience, Self and Identity, DOI: 10.1080/15298868.2021.1933582
To link to this article: https://doi.org/10.1080/15298868.2021.1933582
© 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
Published online: 07 Jun 2021.
Submit your article to this journal
View related articles
View Crossmark data
Playing to the gallery: investigating the normative
explanation of ingroup favoritism by testing the impact of imagined audience
Vincenzo Iacoviello a,b and Russell Spears b
aUniversity of Geneva, Geneva Switzerland; bUniversity of Groningen, Groningen The Netherlands
ABSTRACT
The present research examined the role of social norms as a determining source of ingroup favoritism in minimal groups.
Across three studies (total N = 814), results showed that ingroup favoritism was reduced when participants imagined the reaction of an external (and egalitarian) entity, as compared to a control con- dition or a condition in which they were explicitly asked to imagine the reaction of ingroup members. In line with the prediction that the desire to appear as a good group member drives conformity to the ingroup norm, the findings also revealed that favoring the ingroup resulted in higher self-esteem (Study 2). This was however limited to situations where the ingroup norm was inferred or induced to be pro-discriminatory, but not when it was anti- discriminatory (Study 3). The proposed explanation is discussed in the light of dominant explanations of ingroup favoritism.
ARTICLE HISTORY Received 24 August 2020 Accepted 18 May 2021 KEYWORDS
Ingroup favoritism; norms;
imagined audience; minimal groups; self-esteem
There is no separation between real and imaginary persons; indeed, to be imagined is to become real, in a social sense [. . .]
- Cooley (1902/1922, p.49)
Understanding intergroup discrimination and prejudice is one of the major challenges of social psychology. Despite the variety of explanations, approaches focusing on norma- tive mechanisms are quite rare. Most often, theoretical models emphasize intra-individual, psychological, factors as the determining sources of intergroup discrimination. A glance at the SAGE Handbook of prejudice, stereotyping and discrimination (Dovidio et al., 2010) indeed reveals the predominance of models focusing on cognitive, affective, and motiva- tional factors. In the few normative perspectives, social norms are usually restricted to a moderating role, where they are considered as releasing or constraining discrimination determined by some other basic process (Crandall & Eshleman, 2003; Ford & Ferguson, 2004; Jetten et al., 1996). The present paper aims at going one step further, by exploring the role of social norms as a determining source of intergroup discrimination in minimal groups.
CONTACT Vincenzo Iacoviello Vincenzo.Iacoviello@unige.ch University of Geneva, Switzerland The authors have no conflict of interest to declare.
https://doi.org/10.1080/15298868.2021.1933582
© 2021 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis Group.
This is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any med- ium, provided the original work is properly cited, and is not altered, transformed, or built upon in any way.
A Normative Perspective on Intergroup Discrimination
Although normative accounts of intergroup discrimination currently tend to be out of vogue, they were more common many decades ago (see Horowitz, 1936; Minard, 1952;
Pettigrew, 1958; Sherif & Sherif, 1953). In the 60s-70s, Tajfel and collaborators (e.g., Billig &
Tajfel, 1973; Tajfel, 1970; Tajfel et al., 1971) provided new insights into discrimination dynamics, by showing that people discriminate in minimal conditions, that is, when they have no knowledge about or direct investment in the intergroup context. At first sight this phenomenon would seem to have very little to do with social norms because these minimal groups were entirely new and unknown. This notwithstanding, Tajfel originally saw the origins of ingroup favoritism precisely in terms of social norms (e.g., Tajfel, 1970).
Specifically, he argued that (1) people learn through their social interactions that favoring ingroup members is the prevalent norm in most intergroup contexts, and that (2) this perception is then generalized to the minimal situation, in which the ingroup norm is perceived, other things being equal, as being discriminatory by expectation and thus by default.
However, before testing it, this explanation was superseded by the account outlined in social identity theory (SIT), which focused on ingroup favoritism addressing the need for a positive and distinctive social identity (Tajfel & Turner, 1979). As part of this theoretical shift, normative explanations were criticized for two main reasons (see Pettigrew, 1991;
Turner, 1980). The first issue concerned circular reasoning. Broadly speaking, the norma- tive assumption is that people discriminate because others discriminate, which simply raises the question of why others discriminate (leading to an infinite regress). Hence, this approach was accused of simply re-describing, rather than explaining ingroup favoritism.
The second issue concerned the normative complexity of the social world. For instance, in addition to the discriminatory norm, scholars also acknowledged the prevalence of a fairness norm. Accordingly, assuming that discrimination is the default norm would be an over-simplification of social reality.
We argue that these concerns are not as severe as previously assumed, and in any case should not have led to the complete dismissal of Tajfel’s normative perspective. We believe that the circularity issue is relevant when referring to descriptive norms (i.e., simply what others do), but less so for injunctive norms (i.e., what others encourage us to do). This distinction was not considered at the time because it was only theorized and popularized about 20 years later (see Cialdini et al., 1990). In order to address the normative complexity issue, we also suggest that discrimination and fairness are indeed both normative, but that these norms are typically promoted by two different social agents. On the one hand, ingroup favoritism is clearly perceived as being promoted by the ingroup and its members (e.g., Assilaméhou & Testé, 2013; DeLamater et al., 1969). On the other hand, fairness norms are generally originally promoted by external entities, which are often supra-ordinate to (or “above”) the intergroup situation and act as “moral referees” of the relationship between ingroups and outgroups. Indeed, external bodies, such as international organizations (Ishay, 2020) and high-level education (Pascarella et al., 1988) tend to promote humanitarian values where all human beings should be treated equally (see also Mcfarland et al., 2019).
To address these concerns and properly test Tajfel’s abandoned assumptions, we believed it necessary to show that (1) people perceive the ingroup (but not supra-
ordinate bodies) to promote ingroup favoritism, and (2) that this perception is responsible for their tendency to favor the ingroup over the outgroup. The first postulate received support from Iacoviello and Spears (2018) experiments. In both a naturalistic setting (i.e., a national context) and a minimal group paradigm, they indeed showed that ingroup favoritism was perceived as being promoted by the ingroup. On the other hand, fairness was perceived as being promoted by external entities (i.e., the UN in the naturalistic context and social scientists in the minimal group paradigm). Half a century after Tajfel’s original hypothesis, the foundations for the normative account of discrimination were established, going beyond models that have focused on normative encouragements or constraints on discrimination.
In the present research, we aim at going one step further in reviving this normative hypothesis, and testing the second postulate, namely that the perceived ingroup norm causes ingroup favoritism. More specifically, we should demonstrate that participants in minimal groups tend to conform to the inferred ingroup norm (rather than to external norms), and examine why they do so. This would provide strong evidence in favor of the general hypothesis that ingroup norm plays a determining role in the emergence of ingroup favoritism. That being said, we neither suggest that the normative explanation should replace the social identity one (inter alia), nor that it is the more compelling one.
We simply argue that it should be considered as one viable account of ingroup favoritism (see also Spears & Otten, 2017).
The Self in the Eyes of (Imagined) Others
Tajfel’s original assumption predicted that the perceived ingroup pro-discriminatory norm would drive participants’ discriminatory behavior in minimal group settings. Research indeed suggests that people conform to ingroup normative expectations (and sometimes internalize them) in order to be fully accepted as group members and avoid social punishments (e.g., Horne, 2009; Rimal & Real, 2005; Sherif & Sherif, 1953). However, participants in minimal groups typically do not anticipate any interaction with other group members, and thus would not face social rewards or punishments. So why would people comply with perceived normative pressures (Deutsch & Gerard, 1955) aimed at favoring the ingroup? A possible answer can be found in Cooley’s (1902/1922) work and the symbolic interactionism current that followed from it (e.g., Blumer, 1969; Mead, 1934;
Shibutani, 1961, see Charon, 1992, for a review). The basic premise of this approach is that people’s behaviors and self-perceptions are shaped by social interactions and the social audiences they imply. These interactions can be real or imagined. In this latter case, people typically think about how their manners, opinions, and behaviors are appraised by others, and this perceived judgment is then reflected in their self-image and thus self-esteem. To quote Cooley (1902/1922): “ . . . so in imagination we perceive in another’s mind some thought of our appearance, manners, aims, deeds, character, friends, and so on, and are variously affected by it.” (p.93). Therefore, people adjust their behavior according to what they believe is socially approved and praised, especially by significant others (e.g., Cooley, 1902/1922; Mead, 1934) and members of their reference groups (Shibutani, 1955).
Accordingly, behaviors are not just driven by instrumental motives, but also by more symbolic ones. People not only try to maximize the materialistic benefits that stem from social approval, but they are also (perhaps mainly) motivated to appear as a good person
and achieve a positive image of themselves in the eyes of important others, generating a positive self-esteem.
In line with this view, Leary’s sociometer theory (Leary, 2005; Leary & Baumeister, 2000) defines self-esteem as an indicator of people’s perceived value in a social network.
Moreover, enhancing one’s self-esteem does not necessarily require actual or direct approval and admiration from peers (or ingroup members), but also results from imagin- ing these positive reactions (Marwick & Boyd, 2011; Tice, 1992). Therefore, we argue that people conform to social expectations, because the imagined social outcome resulting from this behavior is likely to garner imagined approval. Imagining that ingroup members praise them for being loyal group members and showing ingroup favoritism should thus boost self-esteem.
In sum, symbolic interactionism posits that social approval derived from imagined ingroup audiences could be just as powerful as the impact of real or co-present audiences for a further important reason. As self-categorization theory (SCT; Turner et al., 1987) makes clear, true group influence does not come from an external “group pressure”
(Deutsch & Gerard, 1955), but emanates from the internalized social identity. As a consequence, influence does not need co-presence or surveillance by the group to realize or reenforce it (Turner, 1991). Moreover, the social identity model of deindividua- tion effects has argued that the physical absence of other ingroup members can even paradoxically lead to stronger group influence effects because group identity often becomes more salient in the absence of individuals or individuating information in situ (Postmes & Spears, 1998; Spears, 2017, 2021). In the minimal group paradigm, group members typically do not see who is and who is not a member of their group, arguably increasing the salience and impact of group identity (Spears et al., 2009; Tajfel, 1978).
Ingroup Favoritism and Self-Esteem
The literature on Self-Esteem Hypothesis (SEH) relating to group discrimination is parti- cularly relevant for the normative perspective of discrimination, since it addresses the relationship between ingroup favoritism and self-esteem. Indeed, the first corollary of the SEH is that favoring the ingroup should lead to greater self-esteem (see Abrams & Hogg, 1988). This assumption stems from social identity theory’s motive for a positive social identity (see also Oakes & Turner, 1980). Accordingly, ingroup favoritism contributes to positively differentiating the ingroup from a relevant outgroup, thereby enhancing the ingroup status and, by extension, its members’ self-esteem. This assumption has received generally supportive evidence (for reviews, see Martiny & Rubin, 2016; Rubin & Hewstone, 1998) especially when refined to consider group norms (Hertel & Kerr, 2001; Iacoviello et al., 2017; Scheepers et al., 2009). Specifically, evidence shows that ingroup favoritism increases self-esteem, but only when this behavior is congruent with the ingroup norm (i.e., when it is normative). Ingroup favoritism increases self-esteem when the ingroup norm promotes discrimination, but it decreases self-esteem when the ingroup norm promotes fairness. This suggests that self-esteem more likely results from conformity to the inferred ingroup norm (as claimed by the normative perspective), than from the act of favoring the ingroup per se (as claimed by the classical understanding of the SEH). In other words, there is compelling evidence that self-esteem primarily stems from being a good group member, rather than being member of a good group.
To summarize, we hypothesize that in the minimal group paradigm people show ingroup favoritism in order to conform to the injunctive ingroup norm, which is inferred by default as promoting ingroup favoritism. This conformity would appear as a result of people’s desire to enhance their self-esteem by feeling that they are good group members.
The current research
The aim of the present research is to show that, in minimal groups, conformity to the inferred ingroup norm plays a determining role in the emergence of ingroup favorit- ism. In order to test this idea, we examined the impact of an imagined audience on the tendency to favor ingroup members over outgroup members. We argue that people show ingroup favoritism because they imagine other ingroup members to react favorably to this behavior. Ingroup favoritism should thus be reduced when participants are focused on an alternative and egalitarian entity, which is external to the intergroup context. Accordingly, Hypothesis 1 proposes that ingroup favoritism is greater when people infer or imagine the reaction of their fellow ingroup members than when they imagine the reaction of an entity that is external to the intergroup context (in this case, social scientists). In the minimal group paradigm, social scientists might indeed be considered as “moral referees” of the intergroup situation insofar as they are perceived as promoting intergroup fairness (see Iacoviello & Spears, 2018).
Two further hypotheses are aimed at investigating the boundary conditions of this predicted effect. We first examined whether any ingroup favoritism resulting from con- formity to the inferred (discriminatory) ingroup norm is aimed at satisfying people’s motivation to be affiliated to social groups. If this is the case, Hypothesis 1a predicts that the impact of the imagined audience on ingroup favoritism is more prominent when affiliation motives are high (vs. low). Second, as the effect of imagined audience should appear as the result of people conforming to the inferred pro-discriminatory ingroup norm, it should only be observed when the norm promotes ingroup favoritism, but not when it promotes fairness (H1b).
We also investigate why people would conform to the ingroup norm when the likelihood of facing actual social reward or punishment is absent (i.e., as it is the case in the minimal group paradigm). As explained above, conformity stems from people’s desire to enhance their self-image because they believe they have acted as good group members not because of tangible rewards or punishments from the group (cf. Deutsch & Gerard, 1955). Therefore, conforming to the inferred ingroup discriminatory norm by showing ingroup favoritism should result in increased levels of self-esteem. In particular, Hypothesis 2a states that favoring the ingroup enhances self-esteem, because people infer the reaction of ingroup members to be positive. Further, this tendency should only appear when the ingroup norm promotes ingroup favoritism, but not when it promotes fairness (H2b). In this latter condi- tion, ingroup favoritism should rather be associated with decreased self-esteem, because people infer the reaction of ingroup members to be negative.
Three studies tested these hypotheses. Study 1 examined H1 and H1a, Study 2 examined H1, H1a and H2a, and Study 3 investigated H1b and H2b.
All studies have been conducted following ethical guidelines specified in the APA Code of Conduct, and have received approval form the University of Groningen’s Ethical
Committee Psychology (No 16,376-O, No 16,402-O, No 17,001-O). Data can be found on the OSF platform (doi:10.17605/OSF.IO/63,852).
Study 1
The first study manipulated the audience participants were asked to imagine while allocating resources to ingroup and outgroup members. They were either asked to imagine ingroup members or an external entity (i.e., social scientists). We included a control condition, in which they computed the allocation task without any imagined audience. According to our rationale, ingroup favoritism is driven by people’s tendency to conform to the ingroup injunctive norm that is inferred to promote ingroup favoritism by default. Ingroup favoritism should therefore be as evident in the control condition (the
“default” ingroup norm assumption) as when participants are asked to imagine the reaction of ingroup members, and should be reduced when they are focused on an alternative source promoting intergroup fairness (i.e., an external entity). Consequently, in line with H1, we predict that ingroup favoritism should be greater in both the ingroup and the control conditions, than in the external entity condition.
To test H1a, we also measured participants’ affiliation motives. At the beginning of the study, participants thus answered a need to belong scale, as well as a “groupiness” scale, measuring the perceived value of (being in) groups. The effect of imagined audience on ingroup favoritism should be more pronounced among people who are highly motivated to be affiliated to social groups.
Method
Participants and design. Participants were recruited on Amazon’s MTurk, which has been shown to be a valid and reliable source of data collection (e.g., Berinsky et al., 2012;
Buhrmester et al., 2011). They were compensated for their time with US$0.35. Following Simmons et al.’s (2011) recommendation, we used the rule of thumb of 50 participants per experimental condition. The final sample consisted of 158 US participants (100 women and 58 men; Mage = 36.91 years, SDage = 12.14). Most of them (97.5%) were American citizens. They were randomly assigned to one of the three experimental conditions (imagined audience: ingroup vs. external entity vs. control).
A sensitivity power analysis using G*Power indicates that this sample size provided 80% power to detect effect sizes of ηp2 = 0.06 or greater (α = 0.05). Considering Richard, Bond and Stokes-Zoota’s (2003) meta-analysis showed that the average effect size for ingroup favoritism is 0.12, the present sample size appears well-powered.
Procedure. We first measured participants’ affiliation motives, which were assessed through need to belong and groupiness. Then participants were assigned to minimal groups (see Tajfel et al., 1971). Participants were informed that they would be presented with five pairs of abstract paintings. In each of these pairs of paintings, one painting had been created by a painter named Dusek and the other one by a painter named Tausig. The participants’ task was to indicate, for each pair, the painting they like the most. After they performed the task, they were informed that previous research had established the existence of two groups of people – the group of people preferring Dusek paintings and the group of people preferring Tausig paintings – and that these groups were similar
in terms of size, average age and gender ratio. They then received feedback about their paintings preference. Actually, all participants were assigned to the Dusek group. After the group assignment, they were asked to divide points between ingroup and outgroup members. The task was performed on six different matrices, which were adapted from Bourhis et al. (1994). They were informed that the points should be considered as “having symbolic value (like cent coins . . .)”. Participants were presented with an example of matrix and learned how to allocate points between two other participants denoted by a number and their group membership: one a Dusek member and the other a Tausig member. Before starting the task, we manipulated the imagined presence of a specific audience (see below). Participants then provided their demographics, including political orientation (from 1 = extremely left wing to 7 = extremely right wing; M = 3.56, SD = 1.76) and were fully debriefed. Correlation with both need to belong, r(158) = −.08, p = .231, and groupiness, r(158) = .01, p = .879, were not significant.
Unless otherwise mentioned, answers to all questions in this study were collected on 7-points scales ranging from 1 (“Completely disagree”) to 7 (“Completely agree”).
Measures of affiliation motives
Need to belong. The need to belong to social groups was measured with a five-item sample of the Leary et al. (2013) scale. Sample items were: “My feelings are easily hurt when I feel that others do not accept me” and “I seldom worry about whether other people care about me” (reversed coded). After recoding, a need to belong score was computed such as a higher score indicates higher levels of need to belong (α = .76, M = 4.46, SD = 1.20).
Groupiness. Because the need to belong scale mainly assesses the group affiliation motivation in preventive terms (i.e., the fear of not being accepted), we added a measure of groupiness, which assesses it in a more promoting terms (i.e., people’s appreciation of being in groups). Groupiness was measured,1,2,3 using a 4-item scale (see Kuppens et al., 2021). Examples of items were: “I like building bonds with members of the same group”
and “I can enjoy spending time with a group of people”. A groupiness score was computed such as a higher score indicates higher levels of groupiness (α = .91, M = 4.69, SD = 1.40). Need to belong and groupiness were positively but only moderately correlated, r(158) = .23, p = .004.
Manipulation of imagined audience. Before performing the point-allocation task, participants were asked to imagine that a certain audience is looking at the way they are distributing points, and to consider what this audience would think of them. In the ingroup condition, the imagined audience were “other members of the ingroup”, while in the external entity condition, the imagined audience were “social scientists working on intergroup relations”. In the control condition, there was no mention of an imagined audience.
Ingroup favoritism. Participants’ ingroup favoritism behavior was assessed by subtracting the average points allocated to ingroup members from the average points allocated to outgroup members (see Diehl, 1990). The score of ingroup favoritism was thus computed such that a positive score indicates ingroup favoritism, a negative score indicates outgroup favoritism, and 0 indicates perfect equality (M = 1.59, SD = 3.50).
Results
Main analyses. In order to adequately test for H1, we first computed two Helmert contrasts with the imagined audience variable. Indeed, omnibus analyses test the general hypothesis that “there is a difference between the experimental conditions” without specifying where this difference lies. Therefore, the null hypothesis may be falsified in several ways. Contrast analyses allow us to fix this issue, as they have greater test power (there is no arbitrary division of the effect of variance; see Furr & Rosenthal, 2003). In the present study, we used Helmert contrasts in order to test for H1, according to which ingroup favoritism should be greater in both the ingroup and the control conditions, than in the external entity condition. This specific hypothesis corresponds to the following statistical contrast, C1: ingroup condition = −1; control condition = −1; external entity condition = 2. In addition, the residual variance must explain only a non-significant part of the hypothesis. Accordingly, the residual effect needs to be not significant. To test the residual effect of the hypothesis we created a second contrast that opposes the control condition to the two other conditions, C2: ingroup condition = 1; control condition = −1;
external entity condition = 0). In other words, C1 should be significant, while C2 should not.
We first performed a linear regression analysis on the ingroup favoritism score, with the two orthogonal contrasts (C1 and C2) as predictors. The analysis first showed that the intercept was positive and significantly different from zero, B = 1.59, t(155) = 5.79, p < .001, 95% CI [1.05, 2.14], ηp2 = .18, which indicated a general tendency to favor ingroup members over outgroup members. It also produced the expected effect of C1, B = −0.42, t(155) = −2.16, p = .032, 95% CI [−0.80, −0.04], ηp2 = .03. As illustrated in Figure 1, ingroup favoritism was greater in the modality combining the ingroup and the control conditions (M = 2.02, SE = 0.34) than in the external entity condition (M = 0.75, SE = 0.48). C2 was however not significant, B = 0.31, t(155) = 0.91, p = .365, 95% CI [−0.36, 0.97], ηp2 = .01, indicating that ingroup favoritism did not differ between the ingroup (M = 2.32, SE = 0.48) and the control condition (M = 1.71, SE = 0.49)1.
-1 0 1 2 3 4
Control Ingroup External entity
Ingroup favoritism
Imagined audience
Figure 1. Ingroup favoritism according to the imagined audience (Study 1). Error bars represent ±1 SE.
In order to examine if the effect of the imagined audience on ingroup favoritism was moderated by participants’ affiliation motives (H1a), we then performed two separate linear regression analyses on the ingroup favoritism score, with C1, C2, one of the two affiliation-motive variables (mean-centered) and their interactions (except those including the two orthogonal contrasts) as predictors. Neither need to belong nor groupiness were reliable moderators of C1, all B ≤ 0.09, ts(152) ≤ 0.52, ps ≥ .605, η2p ≤ 0.012.
Complementary analyses. Post-hoc reflections made us wonder whether political ideologies may play a moderating role. Political ideologies indeed mirror values related to intergroup relations (Jost et al., 2003; Raijman et al., 2003). More specifically, conserva- tives’ values tend to be related to ingroup loyalty, while liberals’ values tend to be related to humanitarianism; Caprara et al. (2006); Graham et al., 2011). We therefore performed a linear regression analysis on the ingroup favoritism score, with C1, C2, political orienta- tion (mean-centered) and their interactions (except those including the two orthogonal contrasts) as predictors. The analysis showed a significant C1 × Political orientation interaction, B = −0.26, t(152) = −2.30, p = .023, 95% CI [−0.49, −0.04], ηp2 = .03.
Investigation of the simple effects showed that, among conservatives (+1 SD), ingroup favoritism was greater in the modality combining the ingroup and the control conditions (M = 2.94, SE = 0.45) than in the external entity condition (M = 0.29, SE = 0.71), B = −0.88, t (152) = −3.06, p = .003, 95% CI [−1.44, −0.31], ηp2 = .06. Conversely, among liberals (−1 SD), ingroup favoritism did not vary according to the imagined audience (Ms = 1.00 and 1.14, SEs = 0.47 and 0.64, for the modality combining the ingroup and the control conditions and the external entity condition respectively), B = 0.04, t(152) = 0.16, p = .872, 95% CI [−0.49, 0.57], ηp2 < .01. The main effect of political orientation was marginally significant, B = 0.28, t(152) = 1.76, p = .081, 95% CI [−0.04, 0.59], ηp2 = .02, and the C2 × political orientation was not significant, B = 0.01, t(152) = 0.06, p = .956, 95% CI [−0.37, 0.39], ηp2 < .013.
A comparison between the basic model and the three moderation models with the detailed effects can be found in Table 1.
Discussion
Consistent with H1, results showed that ingroup favoritism was greater in the control condition and in the imagined presence of ingroup members condition, than when they imagined the presence of an external entity. Findings did not support the assumption that this effect is more pronounced among participants who are highly motivated to be affiliated to social groups (H1a). Complementary analyses, however, revealed that political orientation was an effective moderator of the imagined audience effect. This effect appeared primarily among conservatives. Liberals indeed showed low levels of ingroup favoritism, regardless of the imagined audience.
Study 2
Study 2 first aimed to test H1 and H1a, and replicate Study 1’s findings. As we observed the same outcome in both the control and the ingroup conditions in Study 1, it is likely that the same process – i.e., relying on the inferred ingroup discriminatory norm to allocate points – is at play in these conditions. In order to devise amore parsimonious
design, we therefore only compared the ingroup audience condition to the external entity condition. Moreover, we manipulated the motivation to be affiliated to social groups (in addition to measuring it at the beginning of the study), which would provide a more compelling test of H1a.
We also examined the consequences of showing ingroup favoritism on self-esteem.
H2a predicts that ingroup favoritism should be associated with greater self-esteem.
According to our normative perspective, this effect should be mediated by participants’
inference about the (positive) reaction of their fellow ingroup members. Finally, the present study checked norm perceptions of the ingroup and of the external entity.
Even though previous research has shown that the ingroup norm is perceived as more discriminatory than the social scientists norm (Iacoviello & Spears, 2018), we wanted to acknowledge this in the present study.
Method
Participants and design. Participants were recruited on Amazon’s Mechanical Turk and were compensated with US$0.35. We aimed at recruiting more than 50 participants per cell of the experimental design and therefore recruited about 300 participants. Our final sample size consisted of 308 participants living in the US (192 women and 116 men; Mage
= 36.56 years, SDage = 12.40). All of them were American citizens. They were randomly assigned to one cell of the 2 (imagined audience: ingroup vs. external entity) × 2 (affiliation motive: high vs. low) between-participants design.
A sensitivity power analysis using G*Power suggests that this sample size provided 80% power to detect effect sizes of ηp2 = 0.03 or greater (α = 0.05). Considering that the Table 1. Comparison of the linear regression models in Study 1. Dependent variable is ingroup favoritism.
Model Adjusted
R-Squared B Std. Error t p η2p
1 .022 Intercept 1.593 0.275 5.791 < .001 .178
C1 −0.419 0.194 −2.158 .032 .029
C2 0.307 0.337 0.909 .365 .005
2 .026 Intercept 1.584 0.275 5.769 < .001 .180
C1 −0.430 0.194 −2.217 .028 .031
C2 0.307 0.337 0.913 .363 .005
Belong 0.380 0.230 1.650 .101 .018
C1 × Belong 0.087 0.169 0.519 .605 .002
C2 × Belong 0.263 0.272 0.967 .335 .006
3 .009 Intercept 1.612 0.279 5.784 < .001 .180
C1 −0.434 0.196 −2.211 .029 .031
C2 0.348 0.342 1.015 .312 .007
Groupiness −0.178 0.206 −0.865 .388 .005
C1 × Groupiness 0.039 0.145 0.268 .789 .000
C2 × Groupiness −0.150 0.252 −0.595 .553 .002
4 .058 Intercept 1.547 0.276 5.608 < .001 .171
C1 −0.417 0.193 −2.157 .033 .030
C2 0.072 0.341 0.212 .832 .000
Political 0.279 0.159 1.756 .081 .020
C1 × Political −0.261 0.113 −2.303 .023 .034
C2 × Political 0.011 0.193 0.056 .956 .000
effect size of C1 in Study 1 was of similar magnitude (i.e., ηp2 = 0.03), the present sample size appears well-powered.
Procedure. As in the preceding study, participants answered the need to belong and the groupiness measures, performed the paintings task and were eventually assigned to their minimal group. We then manipulated the motivation to be affiliated to social groups with a task which consisted of reminding them of past events of rejection or inclusion.
Participants were introduced to the point-allocation task and were assigned to one of the two conditions of imagined audience. After they had performed the point-allocation task, they answered measures of the inferred reaction of other ingroup members, of the perceived injunctive norm of the ingroup and the external entity, and of personal self- esteem. Finally, they provided their political orientation (from 1 = extremely left wing to 7 = extremely right wing; M = 3.53, SD = 1.79) and their demographics. Unless otherwise mentioned, answers to all questions in this study were collected on 7-points scales ranging from 1 (“Completely disagree”) to 7 (“Completely agree”).
Measures of affiliation motives
Need to belong and groupiness. Measures of need to belong (α = .78, M = 4.40, SD = 1.26) and groupiness (α = .87, M = 4.91, SD = 1.24) were identical to Study 1’s. The correlation between these two scores was positive and significant, r(308) = .28, p < .001.
Independent variables.
Affiliation motive. The manipulation of the motivation to be affiliated to groups was inspired by Maner et al. (2007; Study 1). Participants were asked to remind themselves of past events in which they had experienced social rejection vs. social acceptance. More specifically, in the rejection condition [acceptance condition in brackets], they were asked to think about moments of their life when they had felt rejected [accepted] by other people or excluded from [included in] a group, and to briefly describe these events in a box below. In line with Maner et al. (2007), participants recalling episodes of social rejection should be more motivated to reconnect with people than participants remind- ing episodes of social acceptance.
Imagined audience. Imagined audience was manipulated in the same way as in Study 1, except that the control condition was absent.
Dependent variables.
Ingroup favoritism. Ingroup favoritism was assessed in the same way as in Study 1.
Inferred reaction of the ingroup. Participants were asked to infer how other ingroup members would react to the way they had allocated the points. Five items assessed a positive reaction: “they would be happy”, “ . . . be satisfied”, “ . . . like me”, “ . . . welcome me”, “ . . . praise me” (α = .92, M = 4.90, SD = 1.33). Six items assessed a negative reaction:
“they would be upset”, “ . . . be disappointed”, “ . . . reject me”, “ . . . avoid me”, “ . . . exclude me”, “ . . . try to convince me to behave otherwise in the future” (α = .92, M = 2.84, SD = 1.39). The score of inferred reaction was computed by subtracting the mean score of positive reaction from the mean score of negative reaction (M = 2.07, SD = 2.48). A positive score thus indicated a relatively positive reaction and a negative score indicated a relatively negative reaction.
Self-esteem. Personal self-esteem was assessed with a five-item sample of the Rosenberg (1979) scale. In order to measure state (rather than trait) personal self-
esteem, we asked to rate to what extent each of the five statements reflected their
“present state of mind”. Sample items were: “I feel that I have a number of good qualities”
and “I feel I do not have much to be proud of” (reversed coding). After appropriate recoding, a self-esteem score was computed such that a higher score indicates greater self-esteem (α = .88, M = 5.41, SD = 1.32).
Norm perception. Each participants (independent of the experimental condition) answered two items measuring both perceptions of the injuctive norm of the ingroup and of the external entity. The ingroup injunctive norm was assessed with the item: “To what extent do you believe that most of the other members of the group Dusek think it is OK to favor their ingroup members over members of the group Tausig?” The external entity’s injunctive norm was assessed with the item: “To what extent do you believe that social scientists think it is OK for people to favor their ingroup members over members of the other group?” They answered on a 7-points scales ranging from 1 (“Not at all”) to 7 (“Completely”).
Descriptive statistics and correlations between all measures are displayed in Table 2.
Results
In order to check that the ingroup norm was perceived as more discriminatory than the external entity norm, we first analyzed norm perceptions. Then, we analyzed results on ingroup favoritism and tested H1 and H1a. Finally, we examined H2a, by analyzing the impact of ingroup favoritism on self-esteem, which should be mediated by the inferred (positive) reaction of the ingroup.
Norm perceptions. We performed a full-factorial repeated-measures ANCOVA on the norm perceptions, with source of the norm (ingroup vs. external entity) as a within- participants factor, and imagined audience (ingroup vs. external entity) and the manip- ulation of affiliation motive (high vs. low) as between-participants factors. As expected, the analysis showed a main effect of the source of the norm, F(1,303) = 31.42, p < .001, ηp2
= .09. The ingroup norm was perceived as promoting ingroup favoritism more than the social scientists norm (Ms = 4.93 and 4.37, SEs = 0.09 and 0.09, respectively). This effect was qualified by an unexpected Source of the norm × Affiliation motive interaction, F (1,303) = 5.30, p = .022, ηp2 = .02, such that the source of the norm discrepancy was greater in the low affiliation motive condition (Ms = 5.11 and 4.32, SEs = 0.12 and 0.14, for the ingroup and the social scientists norms, respectively) than in the high affiliation motive condition (Ms = 4.75 and 4.42, SEs = 0.12 and 0.13, for the ingroup and the social scientists norms, respectively). The analysis also revealed a main effect of imagined Table 2. Descriptive statistics and correlations (Study 2)..
Variable M SD 1 2 3 4 5 6 7 8
1. Need to belong 4.40 1.26 - - - - - - - -
2. Groupiness 4.91 1.24 .28*** - - - - - - -
3. Political orientation 3.53 1.79 .04 .06 - - - - - -
4. Ingroup favoritism 2.11 3.99 .08 .13* .09† - - - - -
5. Inferred reaction 2.07 2.48 .14* .32*** .14* .43*** - - - -
6. Self-esteem 5.41 1.32 −.13* .31*** .02 .13* .37*** - - -
7. Ingroup norm 4.92 1.50 .03 .10† .03 .41*** .12* .10† - -
8. External norm 4.37 1.64 .08 .15* .20** .29*** .14* .07 .38*** -
Notes: *** p < .001, * p < .01, * p < .05, † p < .10
audience, F(1,303) = 3.94, p = .048, ηp2 = .01. Overall, both norms were perceived as promoting ingroup favoritism to a greater extent in the ingroup condition (M = 4.80, SE = 0.11) than in the external entity condition (M = 4.50, SE = 0.10). All other effects were not significant, all Fs(1,303) ≤ 1.82, ps ≥ .178, ηp2 ≤ .01.
Ingroup favoritism. We performed a linear regression analysis on the score of ingroup favoritism, with imagined audience (coded −1 for ingroup and +1 for external entity), affiliation motive (coded −1 for high and +1 for low) and their interaction as predictors.
The analysis first showed that the intercept was positive and significantly different from zero, B = 1.48, t(304) = 4.72, p < .001, 95% CI [0.86, 2.09], ηp2 = .23, suggesting a general tendency toward ingroup favoritism. In line with H1, the main effect of imagined audience was also significant, B = −0.67, t(304) = −2.98, p = .003, 95% CI [−1.11, −0.23], ηp2 = .03.
Ingroup favoritism was greater in the ingroup condition (M = 2.81, SE = 0.32) than in the external entity condition (M = 1.48, SE = 0.31) (see Figure 2). At odds with H1a, the Imagined audience × Affiliation motive interaction was non-significant, B = −0.22, t (304) = −0.96, p = .339, 95% CI [−0.66, 0.23], ηp2 < .01, as was the main effect of affiliation motive, B = 0.30, t(304) = 1.35, p = .179, 95% CI [−0.14, 0.74], ηp2 < .01.
We then performed two linear regression analyses on the ingroup favoritism score, with imagined audience, the manipulation of affiliation motive, one of the two measures of the motivation to be affiliated to groups (mean-centered) and their interactions as predictors. Apart from the previously described main effect of imagined audience, the analysis including groupiness only revealed a marginally significant and positive main effect of groupiness, B = 0.35, t(300) = 1.90, p = .058, 95% CI [−0.01, 0.71], ηp2 = .01. All the interactions including groupiness were non-significant, all Bs ≤ 0.14, ts(300) ≤ 0.75, ps ≥ .454, ηp2 ≤ .01. The analysis with need to belong as predictor showed no effect involving need to belong (neither the main effect nor the interactions), all Bs ≤ 0.21, ts(300) ≤ 1.17, ps ≥ .242, ηp2 ≤ .01. Again, these patterns of results did not support H1a.
As in the previous studies, a complementary analysis investigated political orientation as a potential moderator. We performed a linear regression analyses on the ingroup
-1 0 1 2 3 4
y t i t n e l a n r e t x E p
u o r g n I
Ingroup favoritism
Imagined audience
Figure 2. Ingroup favoritism according to imagined audience (Study 2). Error bars represent ±1 SE.
favoritism score, with imagined audience, affiliation motive, political orientation (mean- centered) and their interactions as predictors. The analysis showed a marginally signifi- cant Imagined audience × Political orientation interaction, B = −0.23, t(300) = −1.86, p = .064, 95% CI [−0.48, 0.01], ηp2 = .01. Consistent with Study 1, examination of the simple effects showed that, among conservatives (+1 SD), ingroup favoritism was greater in the ingroup condition (M = 3.57, SE = 0.44) than in the external entity condition (M = 1.41, SE = 0.45), B = −1.08, t(300) = −3.41, p = .001, 95% CI [−1.70, −0.46], ηp2 = .04.
Among liberals (−1 SD) however, ingroup favoritism was not different according to the imagined audience (Ms = 2.02 and 1.53, SEs = 0.44 and 0.44, in the ingroup and the external entity conditions, respectively), B = −0.25, t(300) = −0.78, p = .437, 95% CI [−0.87, 0.38], ηp2 < .01. All other effects involving political orientation were non-significant, all Bs ≤ 0.20, ts(300) ≤ 1.60, ps ≥ .111, ηp2 ≤ .01.
A comparison between the basic model and the three moderation models with the detailed effects can be found in Table 3.
Self-esteem and inferred reaction of ingroup. Based on a normative account of the self-esteem hypothesis, we examined consequences of discriminating on self-esteem. We first performed a linear regression analysis on self-esteem, with ingroup favoritism as the predictor. Results showed a positive and significant effect of ingroup favoritism, B = 0.18, t (306) = 2.34, p = .020, 95% CI [0.03, 0.32], ηp2 = .02.
Consistent with H2a, we then examined if this effect was accounted for by the inferred positive reaction of ingroup members to their behavior (see Figure 3, left panel). We performed a PROCESS Model 4 mediation analysis, using 10,000 bootstrapped samples following Hayes (2013) recommendations. The model included ingroup favoritism as the predictor, self-esteem as the dependent variable, inferred reaction of the ingroup as the mediator, and imagined audience and affiliation motive as controlled variables. This analysis showed that ingroup favoritism predicted inferred (positive) reaction of the ingroup, B = 0.27 SE = .03, p < .001, 95% CI [0.20, 0.33], which in turn predicted self- esteem, B = 0.20 SE = .03, p < .001, 95% CI [0.14, 0.26]. The direct effect of ingroup favoritism became non-significant when controlling for inferred reaction of the ingroup, B = −0.16 SE = .02, p = .417, 95% CI [−0.05, 0.02]. Finally, the indirect effect, mediated by the inferred reaction of the ingroup, was significant, B = 0.05 SE = .01, 95% CI [0.04, 0.08].
In summary, the act of favoring the ingroup resulted in higher self-esteem, because such a behavior was perceived as being normative (it is inferred to elicit positive reactions from ingroup members).
Discussion
Replicating Study 1, the present findings showed that ingroup favoritism was greater when people imagined that other ingroup members (vs. an external entity) were looking at the way they performed the point-allocation task. Similar to the preceding studies, this effect did not depend on participants’ affiliation motives. Indeed, neither the measures of need to belong and groupiness, nor the manipulation of the affiliation motive moderated the impact of the imagined audience on ingroup favoritism. Given the consistency of the two studies, it seems that conformity to the ingroup pro- discriminatory norm is unlikely to be accounted for by people’s desire to affiliate to groups generally. As Study 1, political orientation appeared to be an effective
moderator. The effect of the imagined audience was more prominent among conserva- tives than among liberals.
Supporting H2a, the results also revealed that ingroup favoritism was followed by enhanced self-esteem. This was mediated by the inferred positive reaction of ingroup members.
Study 3
Study 3 manipulated both imagined audience (ingroup vs. external entity) and the injunctive ingroup norm (pro- vs. anti-discrimination). It first tested H1b, according to which the impact of the imagined audience on ingroup favoritism is dependent on the ingroup norm. Ingroup favoritism should be higher when the imagined audience is the ingroup than when it is the external entity, but only when the ingroup norm is pro-discriminatory (vs. anti-discrimination). The present study also investigated the consequences of discriminating for self-esteem. Based on the normative perspective of the SEH, H2b predicts that favoring the ingroup should only be associated with increased self-esteem, when the ingroup norm is pro-discriminatory. By contrast, when the ingroup norm is anti-discriminatory, ingroup favoritism should decrease self-esteem. Moreover, this effect should be mediated by the inferred reaction of the ingroup. The increased self-esteem in the pro-discriminatory norm condition should appear as the result of perceiving that favoring the ingroup is the expected thing to Table 3. Comparison of the linear regression models in Study 2. Dependent variable is ingroup favoritism.
Model
Adjusted
R-Squared B Std. Error t p η2p
1 .026 Intercept 2.143 0.224 9.551 < .001 .231
Audience -0.668 0.224 -2.976 .003 .028
Affiliation 0.302 0.224 1.348 .179 .006
Audience ×Affiliation -0.215 0.224 -0.957 .339 .003
2 .019 Intercept 2.147 0.226 9.515 < .001 .232
Audience -0.650 0.226 -2.883 .003 .027
Affiliation 0.291 0.226 1.290 .198 .005
Audience ×Affiliation -0.214 0.226 -0.950 .343 .006.
Belong 0.211 0.180 1.173 .242 .001
Audience ×Belong 0.077 0.180 0.426 .670 .000
Affiliation ×Belong -0.023 0.180 -0.130 .897 .003
Audience ×Affiliation ×Belong -0.073 0.180 -0.405 .686 .001
3 .030 Intercept 2.151 0.225 9.563 < .001 .234
Audience -0.635 0.225 -2.826 .005 .026
Affiliation 0.294 0.225 1.308 .192 .012
Audience ×Affiliation -0.201 0.225 -0.894 .372 .006
Groupiness 0.349 0.183 1.903 .058 .002
Audience ×Groupiness 0.138 0.183 0.750 .454 .001
Affiliation ×Groupiness 0.105 0.183 0.571 .568 .003
Audience ×Affiliation ×Groupiness 0.067 0.183 0.365 .715 .000
4 .034. Intercept 2.134 0.224 9.545 < .001 .233
Audience -0.663 0.224 -2.965 .003 .028
Affiliation 0.320 0.224 1.432 .153 .008
Audience ×Affiliation -0.228 0.224 -1.022 .308 .007
Political 0.200 .125 1.600 .111 .011
Audience ×Political -0.232 .125 -1.857 .064 .001
Affiliation ×Political 0.053 .125 0.426 .670 .003
Audience ×Affiliation ×Political 0.016 .125 0.126 .900 .000
do. Conversely, the decreased self-esteem in the anti-discriminatory condition should emerge as the result of perceiving that favoring the ingroup is the wrong thing to do (i.e., fairness is promoted).
Method
Participants and design. Participants were recruited on Amazon’s Mechanical Turk and were compensated with US$0.35. As for the preceding study, we aimed at recruiting about 300 participants. Our final sample size thus consisted of 348 participants living in the US (215 women and 133 men; Mage = 38.61 years, SDage = 12.78). Most of them (97.7%) were American citizens. They were randomly assigned to one cell of the 2 (imagined audience: ingroup vs. external entity) × 2 (ingroup norm: pro-discrimination vs. anti- discrimination) between-participants design.
A sensitivity power analysis using G*Power suggests that this sample size provided 80% power to detect effect sizes of ηp2 = 0.03 or greater (α = 0.05). Considering that the effect sizes of both C1 in Study 1 and the main effect of imagined audience on ingroup favoritism in Study 2 were of similar magnitude (i.e., ηp2 = 0.03 for both effects), the present sample size appears well-powered.
Procedure. Participants performed the painter preference task and were assigned to their minimal group. Before performing the point-allocation task, they were assigned to one of the two conditions of imagined audience and were informed about the injunctive norm of the ingroup. After they had performed the point-allocation task, they answered measures of the inferred reaction of other ingroup members, of the perceived injunctive norm of the ingroup and the external entity, and of personal self-esteem. Finally, they provided their political orientation (from 1 = extremely left wing to 7 = extremely right wing;
M = 3.53, SD = 1.79) and their demographics. Unless otherwise mentioned, answers to all questions in this study were collected on 7-points scales ranging from 1 (“Completely disagree”) to 7 (“Completely agree”).
Independent variables.
Ingroup norm. Participants were informed about the alleged results of “previous research on the group Dusek”. In the pro-discrimination condition [anti-discrimination condition in bracket], they learned that “surveys have consistently shown that the members of the group Dusek think that their ingroup members should [vs. should not]
favor the group Dusek over the group Tausig (that they should allocate most of the points to the group Dusek [vs. that they should allocate about fifty percent of the points to each group]).
Imagined audience. The manipulation of the imagined audience was identical to Studies 2ʹ and 3’s (ingroup vs. external entity).
Dependent variables
Ingroup favoritism. Ingroup favoritism was assessed in the same way as in the preceding studies.
Inferred reaction of the ingroup. A score of inferred reaction was computed by subtracting the mean score of the five positive reactions (α = .92, M = 4.97, SD = 1.32) from the mean score of the six negative reactions (α = .93, M = 2.88, SD = 1.45). A positive
score thus indicated a relatively positive reaction and a negative score indicated a relatively negative reaction (M = 2.09, SD = 2.54).
Norm perceptions. Perception of the injuctive norm of the ingroup and the external entity was measured with the same single items as in Study 2.
Self-esteem. Personal self-esteem was assessed with the same five-item sample of the Rosenberg (1979) scale as in Study 2 (α = .87, M = 5.53, SD = 1.34).
Descriptive statistics and correlations are displayed in Table 4.
Results
We first analyzed norm perceptions and then tested H1b by analyzing results on ingroup favoritism. Finally, we examined the relationship between ingroup favoritism and self-
a
b
Figure 3. A) Mediation model of Study 2. B) Moderated mediation model of Study 3. Inferred reaction = Inferred (positive) reaction of other ingroup members.
